Observations of Atmospheric Dynamics in 3D with LEO-GEO and GEO-GEO

Stereo ImagingJames L. Carr1

28 February 2019

6404 IVY LANE SUITE 333 GREENBELT, MD 20770

1Carr Astronautics, 6404 Ivy Lane, Suite 333, Greenbelt, MD 20770, USA, jcarr@carrastro.comNOAA sponsorship NA14NES4320003 (CICS, UMD)NASA sponsorship NNG17HP01C, TO#28 (SAMDA, SSAI)

Collaborations

• NOAA Collaboration– Jaime Daniels, Houria Madani (Carr Astro),

Wayne Bresky, Jeff Key– Focuses on GEO-GEO combinations– In progress with preliminary results

• NASA Collaboration– Dong Wu, Michael Kelly (APL), Jie Gong– Focuses on LEO-GEO combination– First-year finished, second year starts March 1st

– Starting to work with JPL2

MISR & GOES-R• Motivation• Method• Results• Validation

3

MISR Special Issue of Remote Sensing:

Carr, J.L.; Wu, D.L.; Kelly, M.A.; and Gong, J., “MISR-GOES 3D Winds: Implications for Future LEO-GEO and LEO-LEO Winds”, Remote Sens. 2018, 10(12),1885; https://doi.org/10.3390/rs10121885

3D = Velocity with 3D location of Wind in the atmosphere

Motivation

• MISR & GOES each have different strengths and weaknesses for Wind observations– MISR measures cross-track velocity well and altitude (parallax),

but in-track velocity couples to altitude– GOES-R measures two wind components well, but operational

wind products must infer altitude from IR temperature• MISR & GOES should be better together than each

working alone and therefore solve both problems.• Advanced “Image Navigation and Registration (INR)” with

the new GOES-R series makes using GOES-R with MISR attractive (geo-registration better than ~200 m @ nadir).

4

MISR & GOES-R

• LEO on NASA Terra S/C• Fore & Aft-looking Cameras

– An: nadir looking– Af, Aa: ±26.1°

– B, C, D: oblique viewing• Red Band

– 275 m resolution– 360 km swaths

• Winds: Zong, Davies, Muller & Diner, 2002

• GOES-16 stationed at -75.2°

• GOES-17 stationed at -137.2°

• Advanced Baseline Imager– Full-Disk (5, 10, 15-min. refresh)– CONUS (5-minute refresh)– MESO (30, 60-sec. refresh)

• Red Band– 500 m resolution

• NOAA Operational Winds5

JPL

MISR

GOES-RTerra

Multi-Angle Multi-Temporal

MISR Multi-Angle Imagery

6

MISR Red Band(672 nm)

275 m

• Color Separation shows Disparities between Cameras

– In-track is Parallax + (mostly) V-wind– Cross-track is (mostly) U-Wind

• MISR Wind Challenge is the separation of Parallax from V-Wind

(R,G,B) = (Aa, An, Af)∆t = (45s, 0, -45s)

U (East)

V (North)

SOM Projection over WGS84 Ellipsoid(Blocks 60, 61, 62 on P024 O098797)

GOES Multi-Temporal Imagery

7

ABI Band 2 (B02)(640 nm)

Remapped into MISR SOM Projection

Advanced Baseline Imager (ABI)(R,G,B) = (T0-5min, T0, T0+5min)

275m(Native 500 m)

• Color Separation shows Disparities between Frames

– Pure Atmospheric Motion– No Parallax

• T0 picked close to MISR An Time• CONUS scene used here

Disparity Measurements

8

Time

MISR Track

GOES

MISR

Self

• 40x40 MISR pixel templates• 8 x 8 MISR pixel sampling (2.2 km)• Normalized Cross Correlation• Subpixel resolution by Interpolation

An

Disparities

9

Wind Retrieval Model

10

• MISR An is designated reference (n = 0)

• Solve for states at each site; 𝜖𝜖𝑛𝑛 is a function of

– 3 positions (𝛿𝛿0)– U & V winds– (optionally W wind)– No synchronization

• Two global “Bundle Adjustment” states allow fine adjustment of MISR block to align better with GOES imagery

• Nonlinear, sparse-matrix solution of order 5N+2~104 per MISR block

𝜒𝜒2 = �𝑛𝑛=1

𝑁𝑁

𝜖𝜖𝑛𝑛𝑇𝑇𝑊𝑊𝑛𝑛𝜖𝜖𝑛𝑛Minimize:

𝛿𝛿 𝑡𝑡𝑛𝑛 = 𝛿𝛿0 + �⃗�𝑣 � 𝑡𝑡𝑛𝑛 − 𝑡𝑡0

States Times assignedto looks at 𝒪𝒪

Residual Disparities

11

• A Cameras + GOES-16• Disparities ~15 km• Residuals < 275m

Residuals after Solution

Block 61 on P024 O098797

12

MISR+GOES over CONUS 2018

One Frame of 5-minute GOES CONUS Imagery

MISR Path 024

JointlyRetrieved3D-Windswith 2.2 kmsampling

PBL

A Cameras

13

P012O099670B60:67 2018-09-13T15:46:52.006:2018-09-13T15:49:17.218

OR_ABI-L2-CMIPM1-M3C02_G16_s20182561547506_e20182561547564_c20182561548033.nc

Florence MESO 2018

Low-altitude winds feeding in warm, moist air

One Frame of 30-Second GOES MESO Imagery PBL

A Cameras

14

MISR+GOES Full Disk 2018

M4 (FD every5 minutes)

EmulatedM6 (FD every10 minutes)

EmulatedM3 (FD every15 minutes)

N=14,366 N=7,008 N=3,772

One Frame of M4 5-minute GOES FD Imagery

A Cameras

Comparison to MISR Winds

15

MISR Wind-Height Correlation

16

*Slope = -100 s

*Davies, Horváth, Moroney, Zhang & Zhu, 2007

Mueller, Wu, Horváth, Jovanovic, Muller, Girolamo, Garay, Diner, Moroney & Wanzong, 2017

Comparison to GOES Winds

17

-10 -5 0 5

GOES L2 DMW (m/s)

-10

-8

-6

-4

-2

0

2

4

6

MIS

R+G

OES

3D

Win

d (m

/s)

U Wind=-0.04 m/s=0.56 m/sN=1514

-10 -5 0 5

GOES L2 DMW (m/s)

-8

-6

-4

-2

0

2

4

6V Wind

=0.13 m/s=0.54 m/sN=1514

18

200 400 600 800 1000

3D-Wind Pressure (hPa)

200

400

600

800

1000

GO

ES D

MW

Pre

ssur

e (h

Pa)

P024O098797

B02

B14

=34.8404, =60.235 hPa, N=941

-100 0 100 200

B02 Pressure (hPa)

0

20

40

60

80

100

120

140

2000 6000 10000

3D-Wind Height above MSL (m)

2000

4000

6000

8000

10000

12000

GO

ES D

MW

Hei

ght a

bove

MSL

(m)

=-356.14, =611.1741 m, N=941

-2000 -1000 0 1000

B02 Height (m)

0

20

40

60

80

100

120

140

Comparison with GOES IR Height Assignments

PBL

PBL

Multilayer?

Multilayer?

>> Estimated LEO-GEO 3D-Winds Retrieval Uncertainty

Validation: Clear-Sky Terrain

19

< 200 m

< 0.5 m/s < 0.5 m/s

W Winds

• W-Component is observable according to the math model

20

Pose 1 Pose 2

U V

W

2 Velocity Components Observable from each Pose

• Usually W-wind in small, so we generally constrain it to zero

• Interpretation as a true “wind” requires confirmation, may be– Cloud-top growth/collapse– Artifact of side-looks at cloud

• Apparent quality improves with number of poses

2 x (3 + 2) = 10 > 6

W-Component Retrievals

21-98 -96 -94 -92 -90 -88

Longitude (deg)

20

25

30

35

40

45

Latit

ude

(deg

)

Cameras=9=-0.06 m/s=0.29 m/sN=68619

-3m/s

3m/s

=51.6 m=81.2 m

-400 -200 0 200 400

H (m)

0

500

1000

1500

2000

2500

3000

3500

=0.07 m/s=0.23 m/s

-1 -0.5 0 0.5 1

Vu (m/s)

0

500

1000

1500

2000

2500

3000

3500=-0.02 m/s=0.25 m/s

-1 -0.5 0 0.5 1

Vv (m/s)

0

500

1000

1500

2000

2500

3000

3500

4000

=-0.06 m/s=0.28 m/s

-1 -0.5 0 0.5 1

W-Component (m/s)

0

500

1000

1500

2000

2500

3000

3500

UV vs. UVW RetrievalsW-Component

MISRABCD + 1 GOES

Comparison with MISR A + 1 GOES

22

Madani, H. and J. Carr, “Stereo Cloud Top Height Products for the GOES-R Era”, NOAA Satellite Conference, April 2015.

Madani, H., J. Carr, Andrew Heidinger, and Steve Wanzong, “Inter-Comparisons between Radiometric and Geometric Cloud Top Height Products”, American Geophysical Union, December 2015.

2017 Emerging Technologies Workshop

-20 -10 0 10 20 30 40 50 60

Latitude (degrees)

0

5

10

15

20

Clo

ud H

eigh

t (km

)

CALIPSO Top Layer 1

CALIPSO Base Layer 1

CALIPSO Top Layer 2

CALIPSO Base Layer 2

Stereo

IR&D GEO-GEO Stereo ImagingGOES-13, -14, -15

GOES-13, -15, -16

Sandy

23

0 100 m/s

12 km

-1 km

Latit

ude

(deg

rees

)

Wind Speed

Wind Height

60

40

20

0

-20

-40

-60

-140 -120 -100 -80 -60 -40 -20Longitude (degrees)

IR&D Full Disk G-13, -16 (Test Slot)17 September 2017 18:00Z No Synchronization

STAR Study• NOAA can do parallax 3D-Winds NOW

• Objective of our present work is to prove this in a way that has a path into operations to provide alternative height assignments for DMWs

• Validations/comparisons will quantify the quality of parallax heights– Comparison with IR height assignments– MISR-GOES winds– Rawinsondes and aircraft wind in situ measurement

24

3D-Wind Coverage with GOES-E,GOES-W

-200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0

Longitude (degrees)

-80

-60

-40

-20

0

20

40

60

80

Latit

ude

(deg

rees

)

Operational Configuration

3D-Wind Coverage with GOES-E,GOES-W,GOES-Test

-200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0

Longitude (degrees)

-80

-60

-40

-20

0

20

40

60

80

Latit

ude

(deg

rees

)

Configuration from our IR&D Study

No synchronization

Trade-off between vertical resolution and coverage

STAR Study Methodology

25

Time

G-17 Remapped into G-16Fixed Grid

G-16in the G-16Fixed Grid

• Remap G-17 into G-16 fixed grid

• Match each triplet using operational winds matching & clustering algorithm

• Ingest matches into MISR/GOES-heritage retrieval model

• Use MISR/GOES-heritage matching for ground-point validation

Can Mix CONUS/MESO/FD

26

CONUS Retrievals with GOES-16, -17 (Test Slot)

B02

PRELIMINARY

CONUS-CONUSMatches from NOAA code

B02 CONUS Validations

27

MISR + GOES Ground Retrievals

PRELIMINARY

IR Full Disk G-16, -17 (W Slot)

28PRELIMINARYFD-FD

IR Ground Retrievals

29We see less ground with this band than with Band 2• 2.55 % of good retrievals are ground for Band 14• 11.25% of good retrievals are ground for Band 2 PRELIMINARY

Full Disk WV 3D-Winds

30

Mean=10212 m

Mean=9540 m

Mean=859 m

Just for Fun!

PRELIMINARY

Did we retrieve the height? TBDFD-FD

31

B08

B09

PRELIMINARY

WV 3D-Wind Comparisons with DMW Products

The Future

32

3D-Wind Coverage with GOES-E,GOES-W,GOES-SA,MSG,Himawari,PACSAT

-200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0

Longitude (degrees)

-80

-60

-40

-20

0

20

40

60

80

Latit

ude

(deg

rees

)

• LEO-GEO Constellations• LEO-LEO Leader-Follower• Cubesat Deployments

• International Partners• Hosted Payloads